Crypto BI

Overview

The crypto business-intelligence platform I built at Wallex, and my most heavily-authored codebase: a .NET analytics backend organized around Microsoft Orleans grains, two Nuxt front-ends (an internal dashboard and a public UI), and a Python service layer that ingests Telegram signal channels via Telethon.

The backend separates a public REST API from a dedicated BI query API, with command/query handlers on both paths. Telegram parser grains normalize channel chatter into structured signals; everything ships to Kubernetes via the manifests and Helm charts kept in-repo.

• crypto.NET solution — API, BI API, domain, Orleans grains
• uiNuxt internal dashboard (ag-Grid, Monaco, OIDC)
• public-uiNuxt public-facing UI
• python-servicesFlask + Telethon ingestion services

Tech stack

Repository structure

The real solution layout — runtime grains and Telegram parsers are first-class projects beside the domain:

src/
├─ Crypto.API                      public REST endpoints
├─ Crypto.BiApi                    BI query API
├─ Crypto.Handlers                 core command/query handlers
├─ Crypto.Bi.Handlers              BI-side handlers
├─ Crypto.Domain                   domain model
├─ Crypto.Infra                    persistence & integrations
├─ Crypto.Runtime.Grains           Orleans actors (per-symbol state)
├─ Crypto.Runtime.Interfaces       grain contracts
└─ Crypto.Runtime.TelegramParsers  Telegram channel ingestion
tests/     domain + grain test projects
kubernetes/  deployment manifests
database_query/

Design patterns

• Actor model (Orleans)

  Crypto.Runtime.Grains holds hot per-symbol/per-channel state as virtual actors, with contracts isolated in Runtime.Interfaces.

• CQRS-style handlers

  Separate Crypto.Handlers and Crypto.Bi.Handlers projects split the write path from BI reads, each behind its own API head.

• Layered architecture

  API → handlers → domain → infra; the domain project has its own test suite.

• Anti-corruption ingestion

  Telethon parsers normalize free-form Telegram signals into domain events before they touch the model.

System design